Comparative study and selection criteria of linear motors

Initially, linear motors have been particularly dedicated to transportation systems. Nowadays, linear motors are meant to replace a system using a rotating motor and a transmission to realize a linear movement. With linear motors the performances increase considerably since the mechanical limitations are removed. This leads to a better precision, a higher acceleration and a higher speed of the moving part. Therefore, direct drives with linear motors are increasingly used in industrial applications although these solutions need often more investment costs. Different linear motor structures and technologies exist. They can be either induction or synchronous motors with a transverse or a longitudinal flux. Furthermore, linear motors may have several topologies. They can be either long or short stator and double or single sided. All these variants may be combined giving therefore numerous possibilities to perform a linear movement. Hence, to make the best choice for a given application, a global methodology based on the comparison of optimized motors is presented in the thesis. This design methodology is based on figure of merits which are bound to the specifications of the studied application. This method differs from a conventional one since optimized motors with the same objective function are compared. The optimized motors are obtained by an indirect method based on an optimization algorithm (Sequential Quadratic Programming, SQP). An indirect approach differs from the conventional deterministic one for which at least one parameter must be fixed to obtain a motor pre design, since there are constraints and validity domain which are introduced. The proposed methodology can be applied either to rotating or to linear motor design. The use of an optimization program to perform motor designs requires analytical motor models. The models developed in this thesis take into account the thermal behavior of the motors in order to achieve more realistic results. Furthermore, the analytical models of synchronous motors are thoroughly studied leading to several interesting conclusions. They are based on well known algorithms developed for rotating motors. The proposed models are very accurate in comparison with the FEM program, except for a transversal flux linear motor for which the obtained results are not worthwhile enough to be optimized. This is caused by the structure of the motor which is close to a reluctant motor and imposes to model the motor by a lumped magnetic scheme. Moreover, a global analysis of the windings due to the particularity of linear motor to have an even or odd number of poles is presented in the thesis. The methodology proposed in the thesis is successfully used for an innovative application which deals with a multi mobile system for a lift. For this lift, several cabins travel in the same shaft implying linear motors to move autonomously each cabin. First, by comparing the different motor technologies, the best motor type is selected. Afterwards, the motor windings for the selected motors are analyzed and compared in order to find the most adapted one for this application. The motor is finally optimized, leading to a motor design proposal. This motor design takes into account the thermal behavior, the material cost and the electrical characteristic of the linear motor.

Object Classification and Detection in High Dimensional Feature Space

Charles Dubout

Object classification and detection aim at recognizing and localizing objects in real-world images. They are fundamental computer vision problems and a prerequisite for full scene understanding. Their difficulty lies in the large number of possible object positions and the appearance variations of object classes. This thesis improves upon several classical machine learning algorithms, enabling large computational gains in high dimensional feature space. A common trend in machine learning and computer vision research is to go large scale. In particular, the advent of huge datasets mined from the Internet, and the combination of multiple feature sources have considerably broadened the applications of computer vision. Tasks which were thought impossible a few years ago, such as human action recognition or pose estimation, automatic outdoor navigation, etc., now seem within reach. This dissertation is divided into two parts. The first one deals with the efficient training of a classifier or detector based on a large number of feature extractors, outside the control of the learning algorithm, and therefore of unknown suitability to the task at hand. More precisely, this part presents two kinds of strategies to accelerate the training of Boosting algorithms in such a context: (a) a method to better deal with the increasingly common case where features come from multiple sources (e.g. color, shape, texture, etc., in the case of images) and therefore can be partitioned into meaningful subsets; (b) new algorithms which balance at every Boosting iteration the number of weak learners and the number of training examples to look at in order to maximize the expected loss reduction. Experiments in image classification and object recognition on four standard computer vision datasets show that the adaptive techniques we propose outperform both basic sampling and state-of-the-art bandit methods. The second part deals with linear object detectors, currently the most popular class of detection systems, encompassing template matching, deformable part models, poselets, convolutional neural networks (which internally use linear filters), etc. The main bottleneck of many of those systems is the computational cost of the convolutions between the multiple rescalings of the image to process and the linear filters. We make use of properties of the Fourier transform and clever implementation strategies to obtain a speedup factor proportional to the filter size, both while training and at test time. We also introduce a few modifications to the original Deformable Part Model (DPM) of Felzenszwalb et al. improving its detection accuracy. The gains in performance are demonstrated on the well-known Pascal VOC benchmark, where an increase by one order of magnitude in the speed of said convolutions, and an average improvement of 15% in the accuracy of the detector are established.

Programme doctoral en Informatique, Communications et Information2013

Synthesis of Novel Integrated Actuators Powered by Shape Memory Alloys

Sean Thomas

In the modern age of miniaturisation, Smart Materials, a type of material that reacts mechanically to a certain stimulus, have become an integral part of this revolution. Among these materials, Shape Memory Alloys (SMAs), which have the highest volumetric work density, are the ideal candidate in creating lightweight and miniature actuators. These alloys, after being deformed, are able to revert back to their original shape when exposed to heat. This exotic behaviour has allowed them to be the core active component in a plethora of applications such as grippers, bio-mimetic robots and surgical instruments.Despite their high work density, their implementation comes with some challenges. While the Shape Memory Effect (SME), the ability to recover strain when a thermal load is applied, is a remarkable behaviour, it is also a complex and multi-physical one. This complicates their design and makes it difficult to predict their behaviour. Furthermore, these alloys are only able to recover strain when deformed at low temperatures. This implies that a biasing element is required to exploit these materials in reversible actuators. Due to these limitations, the work density of SMAs, when implemented as actuators in robotic systems, are often much lower than their theoretical maximum.In this thesis, the various types of SMA actuator implementations from different applications are examined to understand the design requirements and subsystems that are necessary to build an actuator. A holistic approach is, then, used to construct a design methodology to create highly integrated actuators in the hopes of preventing the work density degradation present in traditional SMA-based systems. Here, in this work, the identified subsystems of the SMA actuator are combined to serve as a multi-functional element in the novel integrated actuator.The work employs different strategies to integrate the SMA actuator into robotic systems while also proposing adapted sizing methodologies. The resulting SMA-powered system can be sized to be lightweight, compact and dynamic. Various case studies are presented that utilise the proposed holistic design approach and sizing methodologies to serve as a proof-of-concept and to validate the methodology.In this work, compliant and flexure-based mechanisms are exploited to exclude the need for a dedicated biasing element and create lightweight SMA-powered grippers to demonstrate the advantages of the proposed methodology. Additionally, utilising the mechanical behaviour of the SME, a lightweight mechanically-controlled crawling robot is designed and implemented. Furthermore, with the help of topology optimization and kirigami-inspired design, the work details the creation of compliant SMA structures that allow the material to generate multiple outputs while remaining compact and easy to assemble. Lastly, a novel bistable gripper is designed and sized to experimentally validate the design methodologies proposed in this work.The work demonstrates the different areas in which the degradation of the work density occur in traditional SMA systems. In this regard, design methodologies accompanied with sizing strategies are proposed that allows the creation of lightweight, high bandwidth and integrated SMA-based robotic systems. The results in this thesis, reveal the extraordinary value of SMAs in creating lightweight robotic systems and presents various strategies to allow the further integration of the alloy within the system.

EPFL2022

A numerical method for the optimal control of switched systems

Switched dynamical systems have shown great utility in modeling a variety of systems. Unfortunately, the determination of a numerical solution for the optimal control of such systems has proven difficult, since it demands optimal mode scheduling. Recently, we constructed an optimization algorithm to calculate a numerical solution to the problem subject to a running and final cost. In this paper, we modify our original approach in three ways to make our algorithm's application more tenable. First, we transform our algorithm to allow it to begin at an infeasible point and still converge to a lower cost feasible point. Second, we incorporate multiple objectives into our cost function, which makes the development of an optimal control in the presence of multiple goals viable. Finally, we extend our approach to penalize the number of hybrid jumps. We also detail the utility of these extensions to our original approach by considering two examples.